Sanjay K. Bose

Adaptive FEC selection for lightpaths in elastic optical networks

We propose a new approach to adaptively select FEC types for lightpaths in elastic optical networks. An ILP model and a spectrum-window-based heuristic algorithm are developed to analyze its performance. The proposed FEC selection scheme can achieve good performance with low FEC overhead.

Optimal Time-Dependent Spectrum Sharing between Neighboring Elastic Optical Channels over a Single Link

Considering the time-dependent bandwidth characteristic of elastic optical channels, we develop an Integer Linear Programming (ILP) optimization model to optimally order a set of elastic optical channels on a single link so as to maximally share optical spectra between neighboring channels for time-dependent traffic demands and minimize the total required spectrum.

Message Forwarding in Sparsely Connected Wireless Networks Using Rateless Codes

Rateless codes can reduce retransmission costs over erasure channels. Existing routing algorithms using rateless codes assume densely connected networks and predecide the forwarder nodes, which subsequently participate in message forwarding. Using a rateless code, which is easy to analyze and implement, we propose practically implementable opportunistic message forwarding algorithms for sparsely connected networks, in which all the nodes can potentially participate in forwarding, and study their performance through simulations.

Adaptive Lightpath FEC Selection in an Optical Network

We propose a new adaptive FEC selection scheme to choose different FEC types for each lightpath based on their OSNRs. An ILP model is developed to evaluate the performance in terms of served lightpath demands and required FEC overheads. The proposed scheme can significantly reduce the required FEC overhead compared to the non-adaptive scheme.

Lightpath blocking analysis for optical networks with ROADM intra-node add-drop contention

The contention factor limits the extent to which lightpaths using the same wavelength can be added/dropped in a Reconfigurable Optical Add/Drop Multiplexer (ROADM) when it is operated in a colorless and directionless fashion. This paper presents an analysis to estimate the probability of blocked lightpath requests when a node of this type is used and validates the results for three traffic models. Simulations confirm the validity of the analytical results for lightpath blocking both for various values of the add/drop contention factor C and for changing load distribution in the network. We observe a saturation trend in the lightpath blocking performance as the add/drop port count per bank increases and that a high enough value of C reduces lightpath blocking to levels obtainable from an ideal, comntentionless ROADM. When C is small, limitations on the add/drop port count per bank is observed to be the dominant cause of blocking lightpaths while intra-node contention effects have only a limited impact. With increasing C, blocking is caused more because sufficient link capacity is not available and not because free add/drop ports are not available摘要阻塞因子限制了相同波长光通道在一个无色、 无向 ROADM 上的分/插。 本文提出了一个分析模型来估算在三种通信模型下此类节点的网络阻塞率。 仿真确认了在各种阻塞因子和不同负载分布情况下, 分析模型对网络光通道阻塞率评估的有效性。 研究表明, 随着每一个分/插池中分/插端口数目的增加, 网络光通道阻塞率下降并最终呈现饱和趋势。 同时, 一个足够大的阻塞因子能有效减少光通道阻塞率, 并能接近于最理想的无阻塞 ROADM 节点的性能。 当阻塞因子较小时, 每个分/插池拥有的分/插端口数目将成为光通道阻塞的主要因素, 而节点内部的阻塞性对整体性能的影响有限。 而随着阻塞因子的增大, 网络中链路可用容量将成为光通道阻塞的主要因素, 可用分/插端口数的影响变小。

Fundamentals of Queueing Networks

In this chapter, we consider service models, where the service to be provided may be represented as a sequence of services provided by several servers. Such a service scenario may be conveniently represented as Networks of Queues as shown in Figures 5.1 and 5.2. In these models, customers/jobs which finish service at a queue may either move on for the next stage of service to another queue (or even re-enter the earlier queue) or may leave the network altogether.

Advanced Queueing Networks

In the previous chapter, we were primarily concerned with queues where the service times were exponentially distributed. Queues with nonexponential service times were not really considered, except in a limited fashion in Section 2.9 while describing the Method of Stages. In this chapter, we consider the detailed analysis of single server queues with infinite buffers where the service times can have any general distribution. The arrival process is still assumed to be Poisson in nature.

Analysis of the M/G/1 Queue in Equilibrium

In the previous chapter, we were primarily concerned with queues where the service times were exponentially distributed. Queues with nonexponential service times were not really considered, except in a limited fashion in Section 2.9 while describing the Method of Stages. In this chapter, we consider the detailed analysis of single server queues with infinite buffers where the service times can have any general distribution. The arrival process is still assumed to be Poisson in nature.

Advanced Queueing Theory

Exact analytical methods are available for studying open and closed queueing networks with product-form solutions. Examples of such networks and the methods for obtaining exact analytical results for them were considered in Chapter 5. These results are applicable only if the queues satisfy all the restrictive conditions that are required on their respective arrival and service processes. These conditions may be quite restrictive and may not allow us to use these techniques under more general conditions.

Simulation Techniques for Queues and Queueing Networks

Exact analytical methods are available for studying open and closed queueing networks with product-form solutions. Examples of such networks and the methods for obtaining exact analytical results for them were considered in Chapter 5. These results are applicable only if the queues satisfy all the restrictive conditions that are required on their respective arrival and service processes. These conditions may be quite restrictive and may not allow us to use these techniques under more general conditions.